⌥
right option — or any key you choose
Press and hold
One dedicated key, system-wide. No window to find, no button to click, nothing to interrupt your flow.
On-device dictation for macOS
Stop typing. Start speaking.
Hold one key and talk. AirVoice turns natural speech into clean text and types it right where your cursor is — in any app, processed entirely on your Mac.
100% on-device
Works in every app
One key, zero clicks
0×
faster than typing
0+
words per minute, speaking
0
bytes of audio leave your Mac
0MB
the entire download
How it works
From thought to text
in one motion.
Press and hold your dictation key, say what you mean, let go. AirVoice does the rest before your hands find the keyboard.
Say it naturally
A quiet pill slides out under your notch with a live waveform. Talk the way you talk — no robot voice required.
Ship it today.
text lands at your cursor
Watch it land
Release the key and your words are typed right where your cursor is. No copy, no paste, no cleanup.
System-wide
Anywhere you can type,
you can speak.
AirVoice follows your cursor, not the other way around. Mail, messengers, notes, docs, code — if there’s a text field, it just works.
Built like a native Mac tool
A serious instrument.
Barely an interface.
Know exactly what
your voice saves
AirVoice keeps score — time saved, dictations, words — all computed locally and rendered in a dashboard that feels like it shipped with macOS.
Every dictation,
still at hand
Recent transcriptions live in the app for three days — copy one again or delete it. Stored on your Mac, nowhere else.
Eight Whisper models.
One perfect fit
From a 75 MB sprinter to a 900 MB perfectionist — download several, switch anytime. AirVoice recommends the right one for your hardware.
Your key,
your call
Right Option out of the box. Prefer Fn, F5 or something exotic? Remap the hold-to-talk key in one click.
Quietly native,
light or dark
Starts silently at login, waits in the background, honors your appearance setting. A Mac app that behaves like one.
Private by architecture
Your voice never
leaves this Mac.
Recognition runs locally through WhisperKit on Apple Silicon. There is no server to trust, because there is no server at all.
- The microphone is live only while you hold the key
- Audio is processed in memory — never uploaded, never stored
- History stays on-device and clears on your command
- Models are files you download, inspect and delete
Questions
Good questions,
short answers.
Completely. Once a speech model is downloaded, everything — recording, recognition, insertion — happens on your Mac. Airplane mode changes nothing.
The multilingual Whisper models handle 90+ languages with automatic detection. English-only variants are available when you want maximum speed for a single language.
AirVoice 1.1.0 is a free download — no account, no subscription, no trial timer. Just a 3.6 MB DMG.
macOS 26 or later on Apple Silicon. The speech models run on the Neural Engine, so dictation stays fast without heating up your Mac.
Two: Microphone, used only while you hold the dictation key, and Accessibility, used to type the recognized text at your cursor. Both are granted in System Settings and can be revoked anytime.
Pick from eight Whisper models — 75 MB to 900 MB. AirVoice recommends the best fit for your hardware, optimizes it for your Mac on first run, and lets you switch or delete models whenever you like.
The fastest keyboard
is your voice.
Download AirVoice and dictate your first message two minutes from now.
Download AirVoice
Free · v1.1.0 · 3.6 MB · macOS 26+ · Apple Silicon